The following fixes the assert in mems_in_disjoint_alias_sets_p trigger.
What the comment says can easily happen by using
attribute((optimize("O3"))) inside a -fno-strict-aliasing TU. Dependent
on luck a global variable may get a alias-set non-zero MEM if expanded
from the -O3 context and thus the assert triggers for uses in a
-fno-strict-aliasing function.
Fixed by ignoring recorded alias-sets for -fno-strict-aliasing like
we do everywhere else.
Another possibility is to not allow changing -fstrict-aliasing
via the optimize attribute (but then the same may happen via
LTO of a -O0 and a -O3 unit and pre-recorded type mems).
Bootstrap & regtest pending on x86_64-unknown-linux-gnu.
Richard.
2014-11-26 Richard Biener <rguent...@suse.de>
PR middle-end/63704
* alias.c (mems_in_disjoint_alias_sets_p): Remove assert
and instead return false when !fstrict-aliasing.
Index: gcc/alias.c
===================================================================
--- gcc/alias.c (revision 218078)
+++ gcc/alias.c (working copy)
@@ -417,17 +417,9 @@ get_alias_set_entry (alias_set_type alia
static inline int
mems_in_disjoint_alias_sets_p (const_rtx mem1, const_rtx mem2)
{
-/* Perform a basic sanity check. Namely, that there are no alias sets
- if we're not using strict aliasing. This helps to catch bugs
- whereby someone uses PUT_CODE, but doesn't clear MEM_ALIAS_SET, or
- where a MEM is allocated in some way other than by the use of
- gen_rtx_MEM, and the MEM_ALIAS_SET is not cleared. If we begin to
- use alias sets to indicate that spilled registers cannot alias each
- other, we might need to remove this check. */
- gcc_assert (flag_strict_aliasing
- || (!MEM_ALIAS_SET (mem1) && !MEM_ALIAS_SET (mem2)));
-
- return ! alias_sets_conflict_p (MEM_ALIAS_SET (mem1), MEM_ALIAS_SET (mem2));
+ return (flag_strict_aliasing
+ && ! alias_sets_conflict_p (MEM_ALIAS_SET (mem1),
+ MEM_ALIAS_SET (mem2)));
}
/* Return true if the first alias set is a subset of the second. */
